基于分段损失的生成对抗网络

doi:10.19678/j.issn.1000-3428.0050529

计算机工程 ›› 2019, Vol. 45 ›› Issue (5): 155-160,168. doi: 10.19678/j.issn.1000-3428.0050529

基于分段损失的生成对抗网络

刘其开¹,姜代红²,李文吉³

1.中国矿业大学信息与控制工程学院,江苏徐州 221116; 2.徐州工程学院信电工程学院,江苏徐州 221111; 3.中国国土资源航空物探遥感中心国土资源部航空地球物理与遥感地质重点实验室,北京 100083

收稿日期:2018-02-27 出版日期:2019-05-15 发布日期:2019-05-15
作者简介:刘其开(1992—),男,硕士,主研方向为深度学习、计算机视觉;姜代红,教授、博士;李文吉,工程师、硕士。
基金资助:
国家自然科学基金(51574232);国土资源部航空地球物理与遥感地质重点实验室航遥青年创新基金(2016YFL02);徐州市科技计划项目(KC16SQ78)。

Generative Adversarial Network Based on Piecewise Loss

LIU Qikai¹,JIANG Daihong²,LI Wenji³

1.School of Information and Control Engineering,China University of Mining and Technology,Xuzhou,Jiangsu 221116,China; 2.School of Information and Electrical Engineering,Xuzhou University of Technology,Xuzhou,Jiangsu 221111,China; 3.Key Laboratory of Airborne Geophysics and Remote Sensing Geology,Ministry of Land and Resources,China Aerospace Geophysical Survey and Remote Sensing Center for Land and Resources,Beijing 100083,China

Received:2018-02-27 Online:2019-05-15 Published:2019-05-15

摘要/Abstract

摘要：

生成对抗网络(GAN)在训练过程中未能有效进行生成器与鉴别器间的同步更新,导致模型训练不稳定并出现模式崩溃的现象。为此,提出一种基于分段损失的生成对抗网络PL-GAN。生成器在不同的训练时期采用不同形式的损失函数,同时引入真实样本与生成样本之间的特征级损失,从而使鉴别器提取的特征更具有鲁棒性。MNIST和CIFAR-10数据集上的实验结果表明,与regular GAN、feature-wise GAN相比,PL-GAN具有更高的分类精度与运行效率。

关键词: 生成对抗网络, 模式崩溃, 特征级损失, 分段损失, 半监督学习

Abstract:

Generative Adversarial Network(GAN) fails to effectively execute the synchronous update between generator and discriminator during training,resulting in unstable model training and mode collapse.To solve this problem,a generative adversarial network PL-GAN based on piecewise loss is proposed.The generator uses different loss functions in different training periods,and introduces the feature-wise loss between the real sample and the generated sample,which makes the feature extracted by the discriminator more robust.Experimental results on MNIST and CIFAR-10 datasets show that PL-GAN has higher classification accuracy and operation efficiency than regular GAN and feature-wise GAN.

Key words: Generative Adversarial Network(GAN), mode collapse, feature-wise loss, piecewise loss, Semi-Supervised Learning(SSL)

中图分类号:

TP18

刘其开,姜代红,李文吉. 基于分段损失的生成对抗网络[J]. 计算机工程, 2019, 45(5): 155-160,168.

LIU Qikai,JIANG Daihong,LI Wenji. Generative Adversarial Network Based on Piecewise Loss[J]. Computer Engineering, 2019, 45(5): 155-160,168.

https://www.ecice06.com/CN/Y2019/V45/I5/155

参考文献

［1］CHAPELLE O,SCHOLKOPF B,ZIEN A.Semi-supervised learning［J］.IEEE Transactions on Neural Networks,2009,20(3):542-546.
［2］LEE D H.Pseudo-label:the simple and efficient semi-supervised learning method for deep neural networks［EB/OL］.［2018-02-01］.http://deeplearning.net/wp-content/uploads/2013/03/pseudo_label_final.pdf.
［3］RASMUS A,BERGLUND M,HONKALA M,et al.Semi-supervised learning with ladder networks ［EB/OL］.［2018-02-01］.http://papers.nips.cc/paper/5947-semi-supervised-learning-with-ladder-networks.pdf.
［4］KINGMA D P,MOHAMED S,REZENDE D J,et al.Semi-supervised learning with deep generative models［EB/OL］.［2018-02-05］.https://arxiv.org/pdf/1406.5298.pdf.
［5］GOODFELLOW I J,POUGET-ABADIE J,MIRZA M,et al.Generative adversarial nets［C］//Proceedings of Inter-national Conference on Neural Information Processing Systems.Cambridge,USA:MIT Press,2014:2672-2680.
［6］MAALE L,SNDERBY C K,SNDERBY S K,et al.Auxiliary deep generative models［EB/OL］.［2018-02-06］.https://arxiv.org/pdf/1602.05473.pdf.
［7］DUMOULIN V,BELGHAZI I,POOLE B,et al.Adver-sarially learned inference［EB/OL］.［2018-02-04］.https://arxiv.org/pdf/1606.00704.pdf.
［8］SPRINGENBERG J T.Unsupervised and semi-supervised learning with categorical generative adversarial networks［J］.Computer Science,2015(6):2321-2330.
［9］SALIMANS T,GOODFELLOW I,ZAREMBA W,et al.Improved techniques for training GANs［C］//Proceedings of NIPS’16.Barcelona,Spain:NIPS,2016:2234-2242.
［10］LI Chongxuan,XU Kun,ZHU Jun,et al.Triple generative adversarial nets［C］//Proceedings of NIPS’17.Barcelona,Spain:NIPS,2017:4091-4101.
［11］王坤峰,苟超,段艳杰,等.生成式对抗网络 GAN的研究进展与展望［J］.自动化学报,2017,43(3):321-332.
［12］ARJOVSKY M,BOTTOU L.Towards principled methods for training generative adversarial networks［EB/OL］.［2018-01-25］.https://leon.bottou.org/publications/pdf/iclr-2017.pdf.
［13］ARJOVSKY M,CHINTALA S,BOTTOU L.Wasserstein GAN［EB/OL］.［2018-02-05］.https://arxiv.org/pdf/1701.07875.pdf.
(下转第168页)
(上接第160页)
［14］METZ L,POOLE B,PFAU D,et al.Unrolled generative adversarial networks［EB/OL］.［2018-02-07］.https://arxiv.org/pdf/1611.02163.pdf.
［15］LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based learning applied to document recognition［J］.Proceedings of the IEEE,1998,86(11):2278-2324.
［16］KRIZHEVSKY A,HINTON G.Learning multiple layers of features from tiny images［J］.Handbook of Systemic Autoimmune Diseases,2009,1(4):32-33.
［17］BASTIEN F,LAMBLIN P,PASCANU R,et al.Theano:new features and speed improvements［EB/OL］.［2018-01-25］.http://export.arxiv.org/pdf/1211.5590.
［18］SALIMANS T,KINGMA D P.Weight normalization:a simple reparameterization to accelerate training of deep neural networks［C］//Proceedings of NIPS’16.Barcelona,Spain:NIPS,2016:901-909.
［19］MIYATO T,MAEDA S,KOYAMA M,et al.Distributional smoothing with virtual adversarial training［EB/OL］.［2018-01-26］.https://arxiv.org/pdf/1507.00677.pdf.
［20］RADFORD A,METZ L,CHINTALA S.Unsupervised representation learning with deep convolutional generative adversarial networks［EB/OL］.［2018-02-10］.https://arxiv.org/pdf/1511.06434.pdf.

[1]	胡庆. 多尺度融合与双输出U-Net网络的行人重识别[J]. 计算机工程, 2024, 50(6): 102-109.
[2]	顾永跟, 高凌轩, 吴小红, 陶杰. 非独立同分布下联邦半监督学习的数据分享研究[J]. 计算机工程, 2024, 50(6): 188-196.
[3]	张慧妍, 梁勇, 兰景宏, 赵强. 基于记忆模块与过滤式生成对抗网络的入侵检测方法[J]. 计算机工程, 2024, 50(6): 197-207.
[4]	李田芳, 普园媛, 赵征鹏, 徐丹, 钱文华. 基于CLIP和双空间自适应归一化的图像翻译[J]. 计算机工程, 2024, 50(5): 229-240.
[5]	刘帅威, 李智, 王国美, 张丽. 基于Transformer和GAN的对抗样本生成算法[J]. 计算机工程, 2024, 50(2): 180-187.
[6]	何银银, 胡静, 陈志泊, 张荣国. 融合门控变换机制和GAN的低光照图像增强方法[J]. 计算机工程, 2024, 50(2): 247-255.
[7]	张美美, 秦品乐, 柴锐, 曾建潮, 翟双姣, 闫俊义, 冯二燕. 面向急性缺血性脑卒中的CT生成MRI算法[J]. 计算机工程, 2024, 50(2): 317-326.
[8]	沈梦强, 于文年, 易黎, 宋南. 基于GAN的全时间尺度语音增强方法[J]. 计算机工程, 2023, 49(6): 115-122,130.
[9]	陈仲磊, 伊鹏, 陈祥, 胡涛. 基于集成学习的系统调用实时异常检测框架[J]. 计算机工程, 2023, 49(6): 162-169,179.
[10]	李培育, 张雅丽. 基于改进SRGAN模型的人脸图像超分辨率重建[J]. 计算机工程, 2023, 49(4): 199-205.
[11]	罗嗣卿, 陈慧. 基于生成对抗网络的图像场景转换[J]. 计算机工程, 2023, 49(4): 217-225.
[12]	翟社平, 张宇航, 柏晓夏. 融合实体邻域信息的知识图谱嵌入负采样方法[J]. 计算机工程, 2023, 49(3): 95-104.
[13]	席荣康, 蔡满春, 芦天亮. 基于数据增强与流数据处理的Tor流量分析模型[J]. 计算机工程, 2023, 49(3): 177-184.
[14]	陈锦生, 马文臻, 方少峰, 邹自明. 基于地基气辉图像的大气重力波目标识别[J]. 计算机工程, 2023, 49(11): 13-23.
[15]	李晨, 侯进, 李金彪, 陈子锐. 基于注意力与残差级联的红外与可见光图像融合方法[J]. 计算机工程, 2022, 48(7): 234-240.

选择文件类型/文献管理软件名称

选择包含的内容

基于分段损失的生成对抗网络

Generative Adversarial Network Based on Piecewise Loss

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

基于分段损失的生成对抗网络

Generative Adversarial Network Based on Piecewise Loss

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价